Second-line drug resistance markers as proxy indicators of time to sputum culture conversion among second-line drug resistant tuberculosis patients tested in Uganda: A cross-sectional study

Study design and setting

This was a cross-sectional study conducted on secondary data of SLD resistant tuberculosis patients bacteriologically confirmed between the months of 01/06/2017 and 31/12/2019. The study was done at the Uganda National Tuberculosis Reference Laboratory, Butabika, Kampala, Uganda.

Study population

A total number of 72 SLD resistant tuberculosis samples bacteriologically confirmed and interpreted with second-line line probe assay (LPA) mutation profiles between the months of 01/06/2017 and 31/12/2019 were enrolled in the study. The SLD resistant samples selected were previously detected using second-line LPA and cross-confirmed using Mycobacteria growth indicator tube (MGIT) second-line DST kit. Sample entries with complete SLD mutation profiles and culture data were included in the study. All samples with missing data were excluded from the study.

Study variables

This study was composed of dependent, and independent variables. The independent variables included the TB gene mutations, clinical and demographic factors namely, age, gender, patient category, nationality, patient history, second-line drug resistance status, time to sputum culture conversion. The dependent variables included fluoroquinolone (FQ) resistance or injectable agent resistance status [high-level resistance, low-level resistance, mutation characterization, number of mutations per strain (single mutations, two mutations, three mutations)].

These study variables were measured using descriptive, univariate cox proportional hazard model analyses with the use of the Kaplan-Meier survival curves.as a means of checking whether the variables are related; with the level of significance set at 5%, where a p-value of ≤0.05 was considered statistically significant. Hazard ratios were computed to measure the relationship between the dependent variables found to have strong or weak correlation with the independent variables (time to sputum culture conversion), further described in the statistical analysis section of this paper.

Sample processing, culture and drug susceptibility testing at the NTRL, Uganda

Except for enrolling SLD resistant participants, all procedures were performed according to standard procedures as described from the manufacturer^26,27 with further details as follows; all samples were previously transported to NTRL using the Uganda National TB specimen transport system while maintaining a cold chain. Prior to sample processing, samples were checked upon receipt to ensure they met the minimal sample acceptance criteria N-Acetyl-L-Cysteine-Sodium hydroxide (NALC-NaOH; final concentration of 1.5% NaOH) was used to prepare the samples. The samples were then inoculated onto MGIT and Lowenstein Jensen (LJ) medium. The first and SLDs' MGIT DSTs were carried out in accordance with standard operating procedure.²⁸ The GenoLyse^® kit was used to extract DNA for second line LPA DST in the biosafety level (BSL) 3; the Genotype MTBDRsl V2.0 assay (direct) on the processed sediments or (indirect) on the corresponding positive cultures of the eligible participants; and the results were interpreted in accordance with standard procedures^26,27 (Figure 1).

Figure 1. Flow chart for sample processing and enrolment into the study.

All work involving the manipulation of live cultures, including those from eligible drug resistant samples that initially returned indeterminate results for second line drugs was performed in a fully functional BSL 3 laboratory to ensure safety to personnel and the environment. All waste generated during the laboratory related activities was managed according to the internal safety procedures at the laboratory setting.

Laboratory quality control

All quality control procedures for the culture media preparation, sample processing, culture, SL-DST, LPA (GenoType^® MTBDRsl) and data analysis were conducted according to the NTRL internal quality control/quality assurance procedures which are briefly described as follows; Acceptable internal quality control (IQC) findings were evaluated for the reviewed samples during each run. On the MGIT960 System, comparability testing was carried out using the second line DST kit (Becton Dickinson, Franklin Lakes, New Jersey). LPA testing was also carried out on well-characterized proficiency testing strains from the Supranational Reference Laboratory, Antwerp, Belgium, which were supplied periodically as part of the test's External Quality Assurance requirement. There were records of LOT-to-LOT testing to determine performance for the test reagents and consumables used to test clinical samples. Following the initial evaluation carried out by other independent staff in accordance with the necessary (ISO 15189:2012) international standards,²⁹ a qualified reviewer additionally examined the mutation interpretations after retrieving the data taken into consideration during the study period for medical and testing laboratories. The integrity of the LPA bands was sustained by affixing cello-tape to the strips and placing them against the worksheet, which was then preserved within a sheet protector. With the exception of 10 strips, all of the bands were thus visible to the naked eye.

Data collection methods including sputum culture monitoring

In this study, the NTRL laboratory information systems (LIS) was used to extract data for the MDR-TB patients notified from the TB specimen referral system/network, after which the second line drug susceptibility and sputum culture data from baseline to fifth successive culture were exported in Microsoft Excel file 2016 format. Samples were previously received along with the patient’s forms with demographic information like age, sex, referring diagnostic unit, history of treatment among others. Samples had their information accessioned into the NTRL-LIS upon reception at the NTRL, their request forms were scanned and saved on access-controlled backup folders. Hard copies were also kept in box files. Samples were then processed for mycobacterial culture on Löwenstein-Jensen (LJ) media and drug susceptibility testing. Cultures were previously read daily for the first week and biweekly thereafter for up to two months and any growth was confirmed by its acid fastness. M. tuberculosis complex species identification was done based on colony characteristics, positivity for MPT64 antigen as well as time-to-growth on LJ media. When no growth occurred after four weeks, the sample was said to have undergone conversion to negative and the time taken for the sample to convert was noted in the TB culture register. The time to sputum culture conversion was determined and computed based on the definition by the World Health Organization,³⁰ where the first two consecutive negative sputum culture results were recorded, with a distinct 30-day separation between them. The second line drug resistance mutations data obtained from the GenoType^® MTBDRsl screening was then added to the already collected metadata for the corresponding SLD resistant TB patients.

Statistical analysis

The TB patients whose drug resistance status was previously analyzed following the national diagnostic algorithm³¹ were re-evaluated using their respective second-line line probe assay DNA strips according to the standard procedures from the manufacturer.²⁷ This was followed by the addition of resistance marker data which were manually curated respectively in a protected Microsoft Excel 2016 sheet [Research Resource Identifiers (RRIDs) RRID:SCR 016137]. This sheet contained the patient datasets with corresponding monthly culture conversion time prior to cleaning to standardize mutation curations and subsequently import these results into STATA v15 (RRID:SCR_012763) for analysis. The analysis included descriptive, Univariate cox proportional hazard model analyses with the use of the Kaplan-Meier survival curves. The level of significance was set at 5% and therefore a p-value ≤0.05 was considered statistically significant. The data were presented in the form of summary statistic tables and figures.

The Kaplan-Meier analysis was used for computing the median survival estimates over time in spite of all the non-uniformity in days of sputum sample referrals of the study participants during the course of routine monthly sputum culture. The estimates of the median time to sputum culture conversion across different SLD resistance markers were computed using Kaplan Meier survival curves. The Kaplan Meier survival curves were assessed using Univariate cox proportional hazard model analyses and STATA v15 to determine and compare the distribution of time to sputum culture conversion across the studied second-line drug resistance markers over the whole follow up period per patient. For each time interval, survival probability was calculated as the number of subjects surviving divided by the number of subjects at risk. This yielded the cumulative proportions indicated in the Kaplan-Meier survival estimate curves for the enrolled patients. The hazard ratios were then computed to measure the relationship between the drug resistance markers and time to sputum culture conversion.

Drug-resistance profile

Overall, a total of 20,508 sample entries of new and previously treated patients were screened for eligibility, largely consisting of patients in the MDR-TB follow up 14,880/20,508 (72.5%) patient category (Figure 1). Of the eligible participants, a total of 709 had a final interpretable result of rifampicin and/or isoniazid. Among these was a total of 72 SLD resistant TB patients identified as SLD-resistant using the second-line LPA (GenoType^® MTBDRsl), for whom we provide an account. Of these, 68/72 (94.4%) had resistance to any of the SLD whereas 4/72 had independent eis C14T low-level Kanamycin-resistance conferring mutations. Additionally, 42/72 (58.3%) and 12/72 (16.7%) were classified with resistance to fluoroquinolones only and injectable agents (IA) only, respectively. Among the patients with SLD resistance, 54/72 (75%) were either FQ/IA resistant, while 14/72 (19.4%) were resistant to both FQ and IA while 4/72 (5.6%) had low-level resistance to Kanamycin.

Socio-demographic factors

The samples and data used in this study were mainly from males (80.6%) and almost half of the sample cultures analyzed had been obtained from Ugandan patients (59.7%). The age was a median (interquartile ranges, IQR) of 29 (0–12) years and time to sputum culture conversion of 3 (0–12) months respectively. The proportion of sputum culture conversion was found to be 79.2 % (95% CI: 68.0–87.2) after 18 months of follow up. The distribution of the cultures was as follows that converted between the Ugandan 54.4% 95% CI [27.6, 29.4] and Non-Ugandan patient samples 45.6% 95% CI [27.5, 29.5]. Majority of the cultures that converted had been isolated from males 82.5% 95% CI [27.8, 29.2], see Table 1.

Table 1 above shows the results from studying the variables; sex, age, nationality, patient category, and reason for request with the frequencies of sputum culture conversion. The median time to conversion for all the studied socio-demographic is three months. The p-values for all the studied socio-demographic factors are greater than the statistical significance of 0.05, which implies that none of the studied socio-demographic factors were found to have a statistically significant effect on the time to sputum culture conversion for this studied batch of patients.

Kaplan-Meier survival estimates for time to sputum culture conversion

Out of the 57 sputum cultures that converted, 63%, 86%, 93% and 95% of the converted cultures had converted within 3, 6, 9 and 12 months, respectively (Figure 2).

Figure 2. Kaplan-Meier survival estimate for culture conversion.

In the above figure, the horizontal axis represents the month at which conversion takes place while the vertical represents the cumulative proportion of cultures converting. The median time was achieved at three months across the studied patients (seen when an extrapolation is done at 0.5 from the vertical axis).

Second-line drug resistance conferring gene mutations

The 102 resistance-conferring mutations found among the SLD resistant TB patients previously described by Mujuni et al. (2022) are depicted below in Figure 3.

Figure 3. Distribution of genetic mutations among second-line drug resistant tuberculosis patients.

The highest frequency (70) of fluoroquinolone drug resistance conferring mutations was noted in the gryA locus and least (01) in the gryB E540D locus. The highest frequency (25) of second-line drug resistance conferring mutations was noted in the gryA A90V locus and least (01) in the gryB locus. Predominant mutations (26) conferring high-level resistance to moxifloxacin were found to frequently occur in the gryA D94G locus, and closely trailed by gryA D94N/Y locus (12). The gryA D94N/Y, gryA D94G and gryA D94H mutations, all associated with high-level resistance to moxifloxacin were more frequently observed among MDR-TB treatment follow-up patients 18 (54.55 95% CI: 36.4—72.0) as compared to other patient categories. The highest frequency (20) of resistance to injectable agents was observed in both rrs gene A1401G mutation among the patients. Only one mutation was detected in the gryB E540D locus among them.

Hazard analysis across the second-line drug resistance conferring gene mutations

Even though none of the mutations had a statistically significant hazard ratio, the presence of gryA A90V, gryA D94N/Y, and rrs G1484T mutations showed increased chances of the sputum cultures converting at hazard ratios of 1.37 (95% CI: 0.78–2.40), 1.17 (95% CI: 0.57–2.40), and 1.15 (95% CI: 0.36–3.71), respectively (Table 2).

The rrs G1484T mutation associated with conferring drug resistance to injectable agents was observed to have the shortest median conversion time of 1.5 months, closely trailed by the gryA A90V with 2.5 months. The gryA D94H among the mutations associated with high-level resistance to moxifloxacin was found to have a median conversion time of five months. The hazard ratios across all the evaluated predictors were either >1.0 or <1.0 indicating that survival was better in one of the groups (one with and one without the predictor of interest), much as the p-values do not show a statistical significance to confidently rely on to make such a call. The presence of one or more mutations in the gryA gene locus also did not have any statistical significance on the median conversion time, although a single mutation in this locus showed higher converted proportions (70.8%) than those that had two (8.3%) or three (2.7%) mutations. Even though none of the mutations had a statistically significant hazard ratio, the presence of rrs G1484T, gryA A90V, and gryA D94N/Y mutations showed increased chances of the sputum cultures converting at 1.5, 2.5 and 3 months respectively.

Ethical considerations

This project was presented to the Uganda NTRL Research Committee for ethics approval, and the research committee waived the need since this study was based on secondary data of bacteriologically confirmed SLD resistant tuberculosis patients, thereby granting permission to proceed and conduct the study. The Uganda NTRL research committee, directed by the laboratory manager, granted authorization to access the data and conduct the study, after which the lead author signed a confidentiality agreement and filed it with the data manager in accordance with relevant norms and regulations. All the data collected for the study were fully anonymized and no patient names were used throughout the data collection and analysis phase.

The Uganda National Guidelines for Research involving humans as research participants - (July-2014.page 28) requires that sample collection follows acceptable standard procedures and that any person who collects identifiable human materials shall ensure that appropriate informed consent has been obtained from the sample sources, including consent for storage for possible uses in future.

In our case, the secondary data of bacteriologically confirmed SLD resistant TB patients’ data used in this operational research study were initially collected for routine care between 01/06/2017 and 31/12/2019 in efforts to detect and control the spread of tuberculosis. As a result, patient consent could not be obtained for isolates whose results had already been reported in accordance with the set Ministry of Health and World Health Organization reporting mechanisms and recommendations. Consent (informed consent) to provide a sample does not need to be obtained from the patient whose samples are going to be used in their care and management, according to the current authorized version of the NTRL Clinicians Handbook (Version 4.0, February-2019, Page 7 of 20) issued under the Uganda Ministry of Health.

Additionally, it further states that the patient who voluntarily moves to the laboratory to provide a sample has already consented. As a result, according to the aforementioned national regulations, the need for informed consent is deemed unnecessary according to national regulations and was no longer required.